Assume we have d-dimensional data x where there seems to be strong correlations among dimensions. Principal component analysis (PCA) projects the data to find key components with largest va?